Image Processing Device
Disclosed is an image processing technique which more suitably reduces flicker noise. An image display device is comprised of an input unit which inputs encoded image data containing images encoded by intraframe prediction and interframe prediction, an image decoder which decodes the encoded image data, a filter unit which filters the decoded images output from the image decoder, and a display unit which displays the image after filtering. The filter unit performs at least a process to correct the pixel values in a decoded image decoded by interframe prediction immediately preceding in time a decoded image decoded by intraframe prediction of the decoded images output from the image decoder.
Latest HITACHI CONSUMER ELECTRONICS CO., LTD. Patents:
- Optical information device and optical information processing method
- Optical information recording device and optical information recording method
- Optical information recording/reproducing device, optical information reproducing device, and optical information reproducing method
- Optical information reproduction device and optical information reproduction method
- Light information device and light information processing method
The present invention relates to an image processing technique for decoding encoded image signals to reproduce and display images.
BACKGROUND ARTAs a method for coding moving image data, such as TV signals, with a high degree of efficiency to record or transmit it, MPEG (Moving Picture Experts Group) formats have been developed, and in particular, the MPEG-1 standard, MPEG-2 standard and MPEG-4 standard are acknowledged as global standard coding formats. In addition, the H.264/AVC (Advanced Video Coding) standard and some other standards are adopted as a format for further increasing compression rate.
In general, the high compression efficiency of moving image coding is achieved by predictive coding utilizing correlation of images in a spatial direction or temporal direction, frequency transform, quantization, variable length coding and other processing; however, the above-described coding processes include an irreversible coding scheme that cannot reconstruct the original signals, which yields signal degradation in a decoded image compared with the original image. Such signal degradation caused by the coding processes is hereinafter referred to as coding distortion.
One of the coding distortions is flicker noise that is a degradation phenomenon observed as screen flickering caused by significant changes in luminance and color made per every frame or per a few frames in decoded images. A technique of reducing the flicker noise is disclosed in Patent Literature 1.
CITATION LIST Patent Literature
- PTL 1: Japanese Unexamined Patent Application Publication No. 2006-229411
Patent Literature 1 proposes a technique of reducing flicker noise of decoded moving images that were encoded only by intraframe predictive coding (predictive coding using the correlation of image signals in a single frame); however, this technique cannot always fully reduce the flicker noise of decoded moving images that were encoded by commonly-used interframe predictive coding (predictive coding using the correlation of image signals of a plurality of frames).
The present invention has been made in view of the above-described problem and has an object to more suitably reduce the flicker noise.
Solution to ProblemEmbodiments of the present invention can be structured as cited in the scope of claims, for example,
ADVANTAGEOUS EFFECTS OF INVENTIONAccording to the present invention, flicker noise can be more suitably reduced.
Embodiments of the present invention will be described below.
First EmbodimentThe image display device (100) in
Following is a description about operations of the image display device shown in
In
The filter unit (106) performs a filtering process to correct a decoded image (154) retrieved from the third buffer (105) using the signals of the first decoded image (152) and the second decoded image (153) retrieved from the first buffer (103) and second buffer (104), respectively. The switching unit (107) selects one at a time from the first decoded image (152), the second decoded image (153) and the post-filtered decoded image (155) retrieved from the filter unit (106) to output it as an image signal (156). The selection and output of the image signals (156) by the switching unit (107) are made in the same order as the output order of the decoded images (151) from the image decoder (102). The output unit (108) outputs the image signals (156) output from the switching unit (107) to display them.
A moving image coding method for producing encoded image data to be input to the image display device according to the first embodiment will be now described. Generally, two coding methods are adaptively used in moving image coding: an intraframe predictive coding method using an encoded image within the same frame as a target image to be encoded as a reference image; and an interframe predictive coding method using preceding and following frames on a time axis as reference images. For example, the MPEG-2 standard switches between the two coding methods on a picture-by-picture basis. The picture encoded only by the intraframe predictive coding method is referred to as an I-picture. Hereinafter, the simply represented term “I” denotes an I-picture. Among the pictures encoded by the interframe predictive coding method, the picture that can be encoded from an encoded past picture on the time axis by interframe predictive coding (forward prediction) is referred to as a P-picture. On the other hand, the picture that can be encoded from an encoded future picture on the time axis by interframe predictive coding (backward prediction) and the picture that can be encoded from both encoded past and future pictures by interframe predictive coding (bidirectional prediction) are referred to as B-pictures. Hereinafter, the simply represented terms “P” and “B” denote a P-picture and B-picture, respectively. In consideration of random access or the like executed at decoding and displaying, a data format periodically including I-pictures that can be decoded without reference pictures is generally employed since decoding of the P-pictures and B-pictures requires reference pictures.
Next, flicker noise handled in the embodiment will be described. Since moving images have generally a strong correlation in a temporal direction, the interframe predictive coding method is chosen at a high rate for P-pictures and B-pictures. Thus, distortion generated in the coding process is propagated to P-pictures and B-pictures after an I-picture (from I0 to P8 and from I9 to P17 in
With reference to
In the image coding device (200), an image divider (201) divides an input image (250) into blocks. A subtracter (202) performs a subtraction process to the divided input image blocks (251) and a predicted image block (256), which will be described later, output from a coding mode selecting unit (212) to produce a differential image signal (252). A DCT unit (203) performs discrete cosine transform (Discrete Cosine Transform, hereinafter referred to as DCT) to the differential image signal (252) output from the subtracter (202) on the image block basis to output it to a quantizer (204). The quantizer (204) quantizes the input DCT data to produce quantized data (253). A variable length encoder (205) performs variable length coding to the quantized data (253) to output encoded image data (254) suitable for a transmission line. On the other hand, an inverse quantizer (206) inversely quantizes the quantized data (253) to output inversely-quantized data. An inverse DCT unit (207) performs an inverse DCT to the inversely-quantized data to produce differential block data. An adder (208) adds the differential block data to the predicted image block (256) from the coding mode selection unit (212) to produce a decoded image. A frame memory (209) stores the decoded image to use it as a reference image for an input image to be subsequently encoded. A motion estimation unit (210) performs motion estimation to a reference image (255) retrieved from the frame memory (209) to obtain a predicted image having the minimum difference from the divided input image blocks (251). An intraframe-prediction unit (211) performs intraframe prediction with a reference image retrieved from the frame memory (209) to obtain a predicted image having the minimum difference from the divided input image blocks (251). A coding mode selecting unit (212) selects either one of the predicted images obtained by the motion estimation unit (210) and the intraframe-prediction unit (211) and outputs the selected image as a predicted image block (256) to the subtracter (202). If the coding mode selecting unit (212) selects the predicted image obtained by the motion estimation unit (210), information about the interframe predictive coding method used to obtain the predicted image is output as a coding mode (257) to the variable length encoder (205). If the coding mode selecting unit (212) selects the predicted image obtained by the intraframe-prediction unit (211), information about the intraframe prediction method used to obtain the predicted image is output as a coding mode (257) to the variable length encoder (205). The variable length encoder (205) encodes the quantized data (253) and coding mode (257) with variable length codes and outputs them.
Next, with reference to the above-described
Suppose the pictures in
With reference to
In
First,
In
In addition, the correction filtering process according to the embodiment is to reduce flicker noise caused by temporal changes in image luminance between the reference blocks and correction-filter target block. Therefore,
Next,
Points of difference in
The correction filtering process according to the embodiment is to reduce flicker noise caused by temporal changes in image luminance between the reference blocks and correction-filter target block. In addition, the correction filtering process according to the embodiment can reduce even flicker noise caused by temporal changes in pixel luminance on the moving object. In order to show the most recognizable example of the flicker noise reduction effect on the moving object,
With reference to the examples in
The operations in the flow chart in
The following is a detailed description about steps (S502) and (S503) for extracting reference blocks used for correction in
When the extraction process is applied to the example in
Similarly, when the extraction process is applied to the example in
Furthermore, when the extraction process is applied to the example in
The method for extracting the reference blocks at steps (S502) and (S503) is performed as follows. A method for obtaining a reference block in the image (610) includes calculating sums of absolute differences between the pixels in the target block and the pixels in blocks in the image (610) and selecting a block having the minimum value of the sum of absolute difference.
The above operations will be described below with reference to
Sum of absolute difference=|c00−a00|+|c01−a02|+|c10−a20|+|c11−a22| [Expression 3]
As with Expression 3, the sums of absolute differences between the target block and the block (802), block (803) and block (804) can be calculated. In the four blocks, a block having the minimum sum of absolute difference is regarded as a reference block.
The above-described operations are an example for determining the reference block using the values of existing pixels in the decoded image (610); however, it is possible to obtain an optimal block by, for example, generating pixels at nonexistent pixel points in the reference image (610) by interpolation and using the generated interpolated pixels. In
For example, when a pixel at an intermediate position (½ pixel position) in the horizontal direction between pixels a00 and a02 in
b01=(a00+a02)/2 [Expression 4]
Similarly, interpolated pixel b10 at an intermediate position in the vertical direction between pixels a00 and a20 can be obtained by Expression 5.
b10=(a00+a20)/2 [Expression 5]
Furthermore, pixel b11 at an intermediate position in both the horizontal and vertical directions can be obtained by Expression 6 using the two interpolated pixels b01, b21.
Although the above interpolated pixels are generated at ½ pixel position, more highly accurate interpolated pixels, with ¼, ⅛ or 1/16 pixel accuracy, can be generated by increasing the number of taps. Using thus generated interpolated pixels, the reference block in the decoded image (610) can be obtained with fractional pixel accuracy. For example, the block (805) in
Sum of absolute difference=|c00−b11|+|c01−b13|+|c10−b31|+|c11−b33| [Expression 7]
Evaluation with the fractional pixel accuracy can further improve accuracy compared with evaluation with integer accuracy.
The above description is about an exemplary method for obtaining a reference block in decoded image (610) shown in
The reference block can be obtained by another method. Another example of the methods for determining the reference block will be described below with
In the case of
Similarly, in the case of
In the exemplary operations, for the target block (604) including the background area and track, such as the block (604) in the decoded image (611) in
Determination and extraction of the reference blocks at steps (S502) and (S503) in
Next, a filtering process performed at (S504) in
Similarly, the term “reference block” of the decoded image (610) in the following description denotes the reference block (600) in the decoded image (610) when the process is applied to the example in
Furthermore, the term “reference block” of the decoded image (612) in the following description denotes the reference block (602) in the decoded image (612) when the process is applied to the example of
An exemplary operation of the filtering process shown in (S504) in
Sfil(i)=w0*Sback(i)+w1*Scurr(i)+w2*Sfwd(i) [Expression 1]
(where w0+w1+w2=1)
In Expression 1, Scurr(i) is a value of the i-th pixel in a pre-filtered target block, Sfil(i) is a value of the i-th pixel in a post-filtered target block, Sback(i) is a value of the i-th pixel in a reference block in a decoded image (610), and Sfwd(i) is a value of the i-th pixel in a reference block in a decoded image (612). In addition, w0, w1, w2 are weighting factors for Sback(i), Scurr(i), Sfwd(i), respectively, and their specific settings are shown in Expression 2 as an example.
In Expression 2, T1 denotes an interframe distance between the decoded image (611) and decoded image (610), and T2 denotes an interframe distance between the decoded image (611) and decoded image (612). Gaussian function (
Specifically, the filter unit (106) in
This calculation can realize a filtering process capable of correcting the pixel value of the target block (601) to closely follow the transition of pixel values from the pixel constituting the reference block (600) to the pixel constituting the reference block (602). A detailed correction example of the pixel values will be described later.
Following is a description about operations of the switching unit (107) in
With reference to
Note that the pixel values of each image used in
First,
The blocks (600), (601), (602) are all identical to those in
As to pixels in the static area, such as pixels (630), (6305), (631), (6315), (632), at the same position in the reference block (600), target block (6005), target block (601), target block (6015), reference block (602), respectively, the values of the pixels are listed in
Next,
The blocks (603), (604), (605) are all identical to those in
Pixels in the moving area, such as pixels (633), (6335), (634), (6345), (635), are at positions according to the movement of the moving object; the values of the pixels are listed in
The pixels (633), (6335), (634), (6345), (635) are indicated on a straight line, because the moving object (track (622)) in
First,
Expression 8 is substituted into Expression 2. If P1 is a target image to be filtered, for example, T1=1 (distance between I0 and P1) and T2=8 (distance between P1 and I9) are established, and therefore the substitution of Expression 8 into Expression 2 yields Expression 9.
In addition to the value of w0, Expression 9 shown above can obtain values of w1 and w2, i.e., w1≈0.5065 and w2≈0.0145.
Substituting the obtained values of w0, w1, w2, Sback(i)=13 (decoded image of I0), Scurr(i)=17 (decoded image of P1), Sfwd(i)=30 (decoded image of I9) into Expression 1 yields Expression 10 below, resulting in pixel value of 15 for post-filtered P1.
The results of the above-described processes performed on P1 to P17 are shown in
Thus, the image display device according to the first embodiment of the present invention performs a correction filtering process so that the difference between the values of a pixel on an object of a P-picture immediately before an I-picture and a pixel on an object of the I-picture are reduced, thereby reducing screen flickering.
As described above, the image display device according to the first embodiment of the present invention uses, among decoded images including intraframe-prediction decoded images and interframe-prediction decoded images, the pixels of a first intraframe-prediction decoded image and the pixels of a second intraframe-prediction decoded image following in time the first intraframe-prediction decoded image, to subject the interframe-prediction decoded images located in time between the first intraframe-prediction decoded image and the second intraframe-prediction decoded image to the filtering process with a correction filter.
This smoothes signal variations of the decoded images, which are located in time before and after the intraframe-prediction decoded images as if they sandwich the intraframe-prediction decoded image, in the temporal direction, thereby reducing screen flickering and flicker noise.
Furthermore, the pixel values of the interframe-prediction decoded image are corrected using the intraframe-prediction decoded image whose pixel values are closer to the pixel values of the original image than the pixel values of the interframe-prediction decoded image, thereby reproducing pixel values close to those of the original image and therefore reducing flicker noise.
The image display device according to the first embodiment of the present invention can more suitably reduce flicker noise in the decoded images including intraframe-prediction decoded images and interframe-prediction decoded images.
Second EmbodimentAn image display device according to the second embodiment of the present invention will be described below. The image display device of the embodiment will be described, as with the first embodiment, with reference to the block diagram in
The following description is of the first buffer (103), second buffer (104) and third buffer (105) in
Assuming that the images in
The operations of the switching unit (107) in
With reference to
In the second embodiment, the image (I18) that was encoded by intraframe predictive coding and is located in time after P10 to P17 is an unknown image and therefore is predicted from I0 and I9. In this description, as an example, luminance of I0 to I18 is supposed to vary linearly. Let the interframe distance between I0 and I9 be Tback, the interframe distance between I9 and I18 be Tfwd, the pixel value of decoded image I0 be SI0
With the use of the above-calculated values SI18
The results of the process performed on P10 to P17 will be shown in
Thus, the image display device according to the second embodiment of the present invention also performs a correction filtering process so that the difference between the values of a pixel on an object of a P-picture immediately before an I-picture and a pixel on an object of the I-picture are reduced, thereby reducing screen flickering.
As described above, the image display device according to the second embodiment of the present invention uses, among decoded images including intraframe-prediction decoded images and interframe-prediction decoded images, the pixels of a first intraframe-prediction decoded image and the pixels of a second intraframe-prediction decoded image following in time the first intraframe-prediction decoded image to subject the interframe-prediction decoded image located after in time the second intraframe-prediction decoded image to the filtering process with a correction filter.
The image display device according to the second embodiment of the present invention performs the filtering process to the decoded images following in time the images used as the correction filter. In contrast to the image display device according to the first embodiment that creates a one-GOP delay at maximum to perform the filtering process and requires a buffer for storing decoded image information of the GOP, the image display device according to the second embodiment can mitigate the delay and eliminate the buffer for storing the decoded images.
In short, the image display device according to the second embodiment of the present invention can produce the effect inherent in the image display device of the first embodiment and also can mitigate the delay caused by the filtering process and reduce the number of hardware components, thereby realizing more suitable flicker noise reduction.
Third EmbodimentThe image recording and reproduction device (1000) includes an input unit (1001) inputting an encoded image signal that is encoded by a predetermined coding scheme, a recording/reproducing switching unit (1002) switching between recording/reproduction of contents input from the input unit (1001) and reproduction of images accumulated in a content storage unit, which will be described later, the content storage unit (1004) storing contents when the recording/reproducing switching unit (1002) performs recording operations, an image signal processor (1003) performing image signal processing described in the first or second embodiment to a content input from the input unit (1001) or an encoded image signal, from the content storage unit (1004), the content and encoded image signal being retrieved by the recording/reproducing switching unit (1002), an image output unit (1005) outputting image signals processed in the image signal processor (1003), an audio output unit (1006) outputting audio signals output from the recording/reproducing switching unit (1002), a controller (1007) controlling respective components in the image recording and reproduction device (1000), a user interface (1008) used by a user to operate the image recording and reproduction device (1000), and some other components.
The image recording and reproduction device according to the third embodiment of the present invention causes the image signal processor (1003) to perform the filtering process, described in the first embodiment or second embodiment, to the decoded image signal.
The third embodiment can provide an image recording and reproduction device capable of more suitably reducing flicker noise of decoded images including intraframe-prediction decoded images and interframe-prediction decoded images, which are input into the input unit or stored in the content storage unit and reproduced, and outputting the decoded images.
REFERENCE SIGNS LIST
-
- 100: image display device
- 101: input unit
- 102: image decoder
- 103: first buffer
- 104: second buffer
- 105: third buffer
- 106: filter unit
- 107: switching unit
- 108: output unit
- 150: encoded image data
- 151: decoded image
- 152: decoded image output from first buffer (103)
- 153: decoded image output from second buffer (104)
- 154: decoded image output from third buffer (105)
- 155: post-filtered decoded image
- 156: decoded image output from switching unit (107)
- 200: image coding device
- 201: image divider
- 202: subtracter
- 203: DCT unit
- 204: quantizer
- 205: variable length encoder
- 206: inverse quantizer
- 207: inverse DCT unit
- 208: adder
- 209: frame memory
- 210: motion estimation unit
- 250: input image
- 251: input image block
- 252: differential image signal
- 253: quantized data
- 254: encoded image data
- 255: reference image
- 256: predicted image block
- 257: coding mode
- 301: variable length decoder
- 302: inverse quantizer
- 303: inverse DCT unit
- 304: adder
- 305: frame memory
- 306: motion compensation unit
- 350: encoded image data
- 351: differential block data
- 352: reference image
- 353: decoded image
- 354: coding mode
- 1000: image recording and reproduction device
- 1001: input unit
- 1002: recording/reproducing switching unit
- 1003: image signal processor
- 1004: content storage unit
- 1005: image output unit
- 1006: audio output unit
- 1007: controller
- 1008: user interface.
Claims
1-17. (canceled)
18. An image processing device comprising:
- an input unit inputting encoded image data including images encoded by intraframe prediction and images encoded by interframe prediction;
- an image decoding unit decoding the encoded image data;
- a filter unit filtering decoded images output from the image decoding unit; and
- a display unit displaying the filtered images, wherein
- the filter unit performs a process to correct the value of a pixel in, among the decoded images output from the image decoding unit, a decoded image decoded by interframe prediction by referring to only a decoded image decoded by intraframe prediction.
19. The image processing device according to claim 18, wherein
- the decoded image referred to in the filter unit is a decoded image preceding in time the decoded image to be corrected in the filter unit.
20. The image processing device according to claim 19, wherein
- the decoded image referred to in the filter unit is a decoded image decoded by intraframe prediction and immediately preceding the decoded image to be corrected.
21. The image processing device according to claim 18, wherein
- the decoded image referred to in the filter unit is a decoded image following in time the decoded image to be corrected in the filter unit.
22. The image processing device according to claim 21, wherein
- the decoded image referred to in the filter unit is a decoded image decoded by intraframe prediction and immediately following the decoded image to be corrected.
23. The image processing device according to claim 18, wherein
- the decoded image referred to in the filter unit is a decoded image preceding and following in time the decoded image to be corrected in the filter unit.
24. The image processing device according to claim 23, wherein
- the decoded image referred to in the filter unit is a decoded image decoded by intraframe prediction and immediately preceding and following the decoded image to be corrected.
25. The image processing device according to claim 18, wherein
- the decoded image decoded by intraframe prediction is an I-picture, and
- the decoded image decoded by interframe prediction is a P-picture or a B-picture.
26. The image processing device according to claim 18 further comprising:
- a first buffer and a second buffer storing intraframe-prediction decoded images; and
- a third buffer storing interframe-prediction decoded images, wherein
- the filter unit performs a process to correct a signal value of an interframe-prediction decoded image stored in the third buffer by using signal values of two intraframe-prediction decoded images stored in the first buffer and the second buffer.
27. The image processing device according to claim 18, wherein
- the filter unit corrects the value of a pixel in a pixel block, and
- the pixel block is a pixel block on an object having moved according to the movement of an object in the decoded image referred to in the filter unit or a pixel block located at the same position in a frame as a pixel block at which a pixel of the decoded image referred to in the filter unit is located.
Type: Application
Filed: May 12, 2009
Publication Date: Jun 16, 2011
Applicant: HITACHI CONSUMER ELECTRONICS CO., LTD. (Tokyo)
Inventors: Hiroaki Ito (Kawasaki), Isao Karube (Fujisawa), Masashi Takahashi (Tachikawa), Yuto Komatsu (Tachikawa)
Application Number: 13/058,578
International Classification: H04N 7/26 (20060101);